source code for gfortran.

[Tim Prince]

At 06:39 AM 10/10/2004, Tobias Schlüter wrote:

>Tim Prince wrote:
> > I'm searching http://gcc.gnu.org/ml/gcc-testresults/ for an indication of
> > when the last reasonable success occurred on various platforms.  I don't
> > find results going back to gcc-3.5, where I recall there were more 
> successes.
> >
>
>http://gcc.gnu.org/ml/gcc-testresults/2004-10/msg00459.html doesn't look too
>bad to me. Hopefully the emptyif.f90 problems will be fixed soon, they're due
>to bugs in gcc's middleend. forall_3.f90 and ret_pointer_1.f90 are issues with
>the frontend, which have unfortunately existed for a while already, also back
>when it was still called gcc-3.5.

I suppose there's hope for gfortran on other platforms, if ever gcc 
recovers. However, several gcc testers have excluded gfortran in their 
reports.

There are no reported successful testsuite runs in the October list for 
cygwin, for example, although it was relatively good on 32-bit 
cygwin/Windows a month earlier.  I'm sure several of us have tried to 
build, totally without success, on other platforms.  Maybe there's an 
statement (implicit somewhere?) that the group of supported platforms has 
been pruned?

I've not been able to reproduce the reported degree of success on 
linux-x86-64.  I did compare gfortran with pgf90 and ifort on that 
OS.  Curiously, neither gfortran nor pgf90 was able to run FORALL() with an 
array assignment  (nested loops using different syntax for inner and outer 
loop).  Also of possible interest are the few cases where gfortran runs 
faster than ifort vectorized array assignments.


>gcc-4.0 is gcc-3.5, that is except from ongoing developments, IOW it's the
>same branch of gcc's source tree, only the name changed at some point.

Yes, it was called gcc-3.5 back when it worked on more platforms, but I 
can't find the reports.  I brought this up partly in case John was 
interested in a different platform than you.


Tim Prince